Diversity and the Civic Spirit in British Neighbourhoods: An Investigation with MCDS and EMBES 2010 Data

Abstract

Recently, there has been a proliferation of studies investigating the relationship between diversity and outcomes such as social cohesion and civic mindedness. This article addresses several common problems in this field and, using data for British neighbourhoods, elaborates on the experiences of both white British and ethnic minority respondents. We conclude that, if anything, diversity should be encouraged to cement the integration progress of migrants and foster stronger identification with Britain in the second generation. Deprivation at the neighbourhood level along with individual factors such as fear of crime is a much stronger predictor of deterioration of the civic spirit than diversity. Bridging contacts have the expected strong positive association with cohesion outcomes; and contrary to policy concerns no strong negative impact is observed for associational bonding among minority ingroupers.

Introduction

An increasing number of academic studies have examined the processes of ethnic diversification and their impact upon the cohesiveness of British communities, testing specifically whether it is ethnic heterogeneity or economic disadvantage which taxes most the strength of the social glue (Fieldhouse and Cutts, 2010; Laurence, 2011; Laurence and Heath, 2008; Letki, 2008; Sturgis et al., 2011; Twigg et al., 2010).

A particular theme in the literature is that diversity may lead to what Putnam has termed ‘hunkering down’. That is, diversity is alleged to reduce trust in neighbours and fellow-citizens, whether they are from the same or from a different ethnic group, and leads to a reduction in civic engagement and participation generally. A second theme, which is conceptually distinct, is that diversity may lead to inter-ethnic tensions: thus each group may retreat into distinct ethnic strongholds, ‘apart from each other’, sharing little in common. This has been described as ‘reactive ethnicity’ (Leifer, 1981; Portes, 1999). These are distinct ways in which diversity may undermine social cohesion. It is perfectly possible that diversity might lead to ‘hunkering down’ without generating reactive ethnicity, and vice versa.

In this article, we aim to explore a wider range of outcomes than has been usual in the literature, investigating both subjective and behavioural measures of hunkering down as well as a measure of reactive ethnicity. We also compare the ways in which both the majority and the minority groups respond to ethnic diversity. It is sometimes implicitly assumed that diversity has opposite effects on majority and minorities respectively. Thus ethnic heterogeneity is assumed to have harmful effects on the majority population, leading them to hunker down, while, conversely, ethnic homogeneity is assumed to have harmful effects on the minority population, leading them to withdraw from the wider society and to maintain distinct ethnic identities rather than identifying with the wider British society. In this article therefore we investigate whether or not the effects of diversity are symmetrical. We also explore some of the mechanisms that may lie behind the assumed negative or positive effects of diversity, namely theories of threat and of contact.

Theory and Hypothesis Building

In just a few years, there has been a dramatic rise in academic and policy interest in the possible effects of diversity and the implications of growing neighbourhood heterogeneity for modern community life, integration and social cooperation. The majority of studies have focused on the relationship between diversity and trust, and the results appear rather pessimistic when data from the USA is used (Hero, 2003). Alesina and La Ferrara (2000, 2002) found that trust in general and, more specifically, interpersonal trust is lower in more racially heterogeneous communities in the USA. Stolle et al. (2008), comparing the USA and Canada, observed a strong negative effect of heterogeneity on trust. However, they also found that contact may neutralize, but not reverse, the sign of this effect. In this debate one name is particularly prominent, that of the American political scientist Robert Putnam, who argued (2007) that heterogeneity in neighbourhoods lowers trust both in people from other ethnic groups and in people from one’s own ethnic group. Thus, diversity seems to undermine the social glue and, in Putnam’s words, to push people towards ‘hunkering down’.

In contrast, the evidence from mainland Europe is much more mixed (Delhey and Newton, 2005; Gerritsen and Lubbens, 2010; Gesthuizen et al., 2009; Kesler and Bloemraad, 2010). The British data do not fully conform to the US pattern either. Fieldhouse and Cutts (2010) compared the USA and the UK and suggested that in England, diversity is only one of a number of factors that are equally important in accounting for variations in neighbourhood social cohesion, with economic deprivation playing a major role in the deterioration of civicness. Laurence and Heath (2008) and Letki (2008), looking at different predictors of social cohesion in respectively the 2001 and 2005 Citizenship Surveys, found that when the association between diversity and economic deprivation is accounted for, there is no strong evidence that heterogeneity erodes trust. Still, with British data based on the Citizenship Survey 2005, Laurence (2011) established that rising diversity is associated with lower levels of neighbourhood trust, although people with bridging ties were less likely to have negative sentiments. Sturgis et al. (2011) found no association between ‘moralistic’ or generalized trust and diversity but an association between diversity and what they term ‘strategic’ trust, that is trust in neighbours. Nevertheless, this effect was dwarfed by the powerful effect of economic deprivation. A similar result was established by Twigg et al. (2010) who observed statistically significant correlations between neighbourhood heterogeneity and a latent variable combining indicators of trust and cohesion, but the size of the effect was substantively small compared to the much larger effect of neighbourhood deprivation (Becares et al., 2011, note this too).

How will the present article contribute to the literature? One interesting aspect of the overviewed studies is that they all possess a similar structure of analysis and therefore exhibit similar theoretical and methodological limitations. In this study, using two British datasets with a focus on social capital – the Managing Cultural Diversity Survey (MCDS) and Ethnic Minority British Election Study 2010 (EMBES 2010) – we address four major shortcomings and offer a way in which they can be resolved.

Trust=Cohesion, or?

Previous research has elaborated from every possible angle on the relationship between trust and diversity. Yet social cohesion is a complex phenomenon that is unlikely to be captured by one single measure. The studies described in the previous section show that it matters even what type of trust we look into – generalized or neighbourhood (Laurence, 2011; Sturgis et al., 2011). In this study we examine a wider range of measures relevant to social cohesion and civic engagement. We look at measures of generalized trust, neighbourhood trust, willingness to help neighbours, and civic participation, which constitute different aspects of hunkering down (see Table 1 for detailed explanation of all the variables used in this analysis).¹ Unlike the other listed measures of hunkering down, which are attitudinal, civic participation is a behavioural measure of engagement (Almond and Verba, 1963) which thus provides a tougher test of the negative effects of diversity.

Table 1.

Variable list.

Dependent variables
Generalized trust	0=You can’t be too careful; 1=Most people can be trusted
Neighbourhood trust	From 0=Low and medium level trust to 1=Trusting
Willing to help neighbours (MCDS data only)	Willing to help neighbours: From 0=Low and medium levels to 1=High levels
Weak co-ethnic identity	Own ethnicity important. From 0=Identifies strongly or average with co-ethnics; to 1=Weak identification; (EMBES data distinction is between from feeling British to feeling Asian/Black)
Civic action	Variable constructed on the basis of participation battery – signing petition, joined local group, have written to council, protest, joined crime prevention groups (+ boycott EMBES data only), other form of participation. 0=No; 1=Yes
Independent variables (Base model)
lnage	Natural logarithm of age; mean/sd (MCDS): White: 3.774/0.448 Minorities: 3.555/0.388 mean/sd (EMBES): 3.594/0.037
Gender	0=Male; 1=Female
Marital status	0=Single; 1=Married/Cohabiting; 2=Separated/Divorced/Widowed
Ethnicity	0=White (MCDS data only); 1=Black Caribbean; 2=Black African; 3=Indian; 4=Pakistani; 5=Bangladeshi; 6=Other ethnic group
Generation	0=1st; 1=2nd+
Housing	0=Own; 1=Rent from Council; 2=Rent privately; 3=Other
Education	0=No education; 1=GCSE grades A–C or equivalent; 2=GCE A level or equivalent; 3=Higher Diploma; 4=Higher education
Diversity	Index of Ethnic Fractionalization; mean/sd (MCDS) White: 0.330/0.246 Minorities: 0.661/0.189; mean/sd (EMBES): 0.782/0.217 Minorities live in general in more diverse settings than white British.
Independent variables (Conflict model)
LnIMD	Natural logarithm of the Index of Multiple Deprivation; mean (MCDS) White: 2.852/0.708 Minorities: 3.116/0.637; mean (EMBES): 2.802/1.007
Discrimination (EMBES data only)	0=No; 1=Yes
Economic precariousness (EMBES data only)	From 0=Good economic situation to 1=The economic situation of respondent got a lot worse in recent year
Fear of crime (MCDS data only)	From 0=Not worried to 1=Worried
Independent variables (Contact model)
Bonding contact	1=In organization in which co-ethnics are half or more than half of organization; 0=Everybody else
Bridging contact	From 0=Below the median 1=Median, Above the median
N	MCDS (white British): 809; MCDS (minorities): 700; EMBES: 2750; Cases with ‘no answer’ have been dropped

To the popular measures of civic engagement such as trust, we also add co-ethnic identification. The concept of co-ethnic identification is of considerable interest. Sometimes referred to as ‘reactive ethnicity’ or ‘stronghold identity’, it provides an indicator of the salience of ethnic and cultural boundaries vis-a-vis other ethnic groups (Leifer, 1981; Portes, 1999). In this study we operationalize reactive ethnicity as strength of identification with one’s own ethnic group.

This is not to say that other studies have not examined outcomes other than trust. Twigg et al. (2010) analyse combined measures of cohesion and trust; Letki (2008) focuses on informal help between neighbours which we also examine; Laurence and Heath (2008) and Becares et al. (2011) look into whether the community is one in which people say they get on well together; Fieldhouse and Cutts (2010) look into norms (a scale based on ‘helping neighbours’, ‘sharing the same values’, and trust) and participation (a scale based on group activity and civic activity). No other study, however, overviews such an extensive range of measures and dimensions of cohesion and civic mindedness as we do.

Diversity, Majority and Minority Groups

Unlike previous research, we examine separately the effect of diversity on both majority and minority members – a crucial distinction which very few studies attempt (Fieldhouse and Cutts, 2010, point to differences between whites and minorities both in terms of norms and participation; Becares et al., 2011, distinguish between the effects of proportion of whites versus minority groups and own-group ethnic concentration). This is an important limitation in the literature, particularly because the research question of whether diversity undermines cohesion seems to have been formulated largely with the white majority in mind. Thus, as noted earlier, ethnic heterogeneity seems to be ‘bad’ for the majority but ‘good’ for minorities as long as heterogeneity in the case of the latter entails greater exposure to white British people. In the literature and policy discourse on ethnic minorities (if their cohesion patterns are indeed considered), the focus is on the negative effects of co-ethnic concentration on the integration process (Becares et al., 2011). In this article we therefore investigate whether the effects of diversity are the same for both majority and minority groups alike, instead of disregarding its implications for minorities or examining segregation² as problematic solely for minorities (since homogeneity among the majority appears viewed as the essence of social glue).

Size of the Effects

We pay particular attention to establishing and taking into account the substantive importance of the observed relationships. This is necessary because the effect of a certain predictor can be significant in the narrowly statistical sense but have little substantive importance when it comes to shaping the observed patterns (Twigg et al., 2010). Without interpretation of the size of the effect, researchers often lapse into vague generalizations, or, worse, overestimate or underestimate the role of a particular predictor (Dawkins, 2008), which is the main criticism of Putnam’s work (Sturgis et al., 2011). To give a concrete example, it has been argued that the substantive effect of economic deprivation on the erosion of social cohesion may deserve more attention than that of diversity. Other studies (Fieldhouse and Cutts, 2010; Sturgis et al., 2011) have looked into correcting this problem by employing standardized coefficients. But since many of our independent variables are dummy variables for which standardization makes little sense, we have preferred to use ‘marginal effects’ (me) and describe in percentage points how the change in the response variable is related to changes in the explanatory variables. This gives us a clear and simple way of examining the substantive importance of the effects.

Diversity, Conflict, and Contact

It is commonly asked what processes are involved when we observe a negative (or, more rarely, positive) statistical association between an index of diversity and an indicator of social cohesion (Dawkins, 2008). Frequently, relying only on indices of diversity, we have little information on the basis of which to assess the mechanisms involved (Blau, 1977; Quillian, 1995). A measure of residential diversity tells us nothing about the actual form and nature of inter-group relations in an area (Allport, 1954). We therefore explore theories of conflict and contact and treat them as explanatory mechanisms mediating the relationship between diversity and social cohesion (Alesina and La Ferrara, 2002; Sturgis et al., 2011). Most of the overviewed studies have only attempted this half-way, we will argue: they control for level of neighbourhood deprivation through various different indices. According to Conflict Theory, under conditions of increased diversity, the majority may feel threatened as to their economic and political power and feel that resources to which they are entitled are being usurped by newcomers. Hence they may display greater levels of intolerance (Blalock, 1967) or act to preserve ‘social distance’ from the minority groups (Bobo, 1988).

In addition to the standard measure of economic deprivation used by other scholars, we include subjective measures of potential conflict such as respondents’ self-assessment of personal ‘economic precariousness’ (available only in EMBES), as studies have shown that the respondent’s economic concerns may have significant impact on their socializing and community participation (Graham, 2011). With the MCDS data, we introduce a subjective measure of the respondents’ perception of crime in their area. Fear of crime may well be a major source of ‘hunkering down’, and respondents feeling threatened by crime may exhibit heightened racial awareness (Quillian, 1995). We expect that the effects of threat, especially economic threat, will be asymmetrical, affecting the majority group more than minorities (Bobo, 1988).

According to the ‘contact’ theory developed by Allport (1954), racial prejudice will be reduced as a function of inter-group contact and integration. It should be remembered that diversity in itself merely provides opportunities for contact, and cannot be equated with actual contact. Very few studies examine whether contact has a mediating effect on the relationship between diversity and cohesion as only a limited number of surveys include measures of inter-ethnic or inter-racial contact (Laurence, 2011; Semyonov and Glikman, 2009; Stolle et al., 2008). Semyonov and Glikman (2009) show with the European Social Survey that positive contacts are likely to reduce perception of threat and social distance. With British data, Laurence (2011) finds that bridging contact, measured as proportion of outgrouper friends, at best neutralizes the negative effect of diversity but does not reverse it.

Research Hypotheses

• Hypothesis 1. On the basis of the overviewed studies and policy reports, we expect that, for white British, our civic outcomes (both measures of hunkering down and identification with co-ethnics) are affected negatively by increased diversity. For minorities, diversity is also expected to lead to hunkering down but at the same time may reduce identification with co-ethnics.

• Hypothesis 2. If there is a negative relationship between diversity and civicness among the white British, then this is expected to be mediated in part at least by threat. If there is a positive relationship among minorities, particularly with respect to identification with Britain, then this is expected to be mediated by contact.

• Hypothesis 3. The effect of diversity on social cohesion is conditional on the presence of bridging ties and/or sources of conflict. Some significant moderating effects will thus be observed (exemplified by a series of interactions). For example, both white British and minority respondents who report low levels of bridging ties may be more negatively influenced by a rising percentage of outgroupers in their local area than respondents with medium and high levels of contact.

Data and Methodology

This study makes use of two major datasets with comparable questionnaire structure designed to measure a range of social and political attitudes and administered in 2010. The first is the Managing Cultural Diversity Survey (MCDS). This survey was a random location quota sample of 1609 individuals: 809 white British respondents; and, through an ethnic minority booster sample, 800 minority individuals. The second dataset is the 2010 Ethnic Minority British Election Study (EMBES). The EMBES was a random probability sample that encompasses the major ethnic minority groups in the UK with a sample of 2782 minority individuals.

Geographical Units

A frequent criticism of diversity studies, such as that of Putnam (2007), is that they employ area units of analysis that vary greatly in size and, therefore, in their potential diversity mix, and are thus incomparable (Dawkins, 2008). The MCDS uses Middle Super Output Areas (MSOAs) as the primary sampling unit while the EMBES sample uses Lower Super Output Areas (LSOAs). Compared to wards, MSOAs and LSOAs are felt to be more appropriate for this analysis due to the fact that wards differ greatly in size.

The MCDS sampled 200 neighbourhoods with the average of 7.95 respondents per unit. The EMBES sample consists of data points spread across 582 neighbourhoods (LSOAs being nested within MSOAs) with the average size of the cluster being 5.5 individuals.

Independent Variables

We include a range of potential ‘confounding variables’ that previous research has found to be associated with the outcomes of interest (Laurence and Heath, 2008; Putnam, 2007; Sturgis et al., 2011; Twigg et al., 2010): age, gender, respondent’s marital status, type of accommodation, education, ethnicity, length of residence and generation (for the minority sample only).

Diversity

In this article, we use the index of ethnic fractionalization (EF) which has been developed as a measure of ethno-linguistic diversity (Alesina and La Ferrara, 2005) and is commonly used in the literature. We have tried other indices as well with similar results. The index has a simple interpretation, namely the probability that two randomly selected individuals from a given area will not belong to the same ethnic group; and ranges from 0 to 1. The greater the value of the index, the greater the diversity.

$$\text{FRAC}=1-{\displaystyle \sum _{i=1}^N{\pi }_i^2}={\displaystyle \sum _{i=1}^N{\pi }_i}(1-{\pi }_1)$$

Where π_i is the share of group i over the total of the population.

Conflict Measures

To measure economic deprivation (which we take to be a proxy for threat) within the observed communities, we will use the Index of Multiple Deprivation (IMD), which includes a range of economic, social and housing issues. It is a commonly used measure of neighbourhood economic deprivation (Laurence, 2011; Laurence and Heath, 2008; Sturgis et al., 2011, Twigg et al., 2010).

With the EMBES data, we can further control for whether or not the respondent has been a victim of racial discrimination. Minority members who have experienced discrimination may be quite conservative in their residential choices and feel more comfortable in neighbourhoods surrounded by ingroupers (Krysan et al., 2009), which may make them less trusting and reluctant to pursue greater integration among white British.

Contact Measures

It is easier said than done to construct a meaningful contact measure. Often, contact measures are based on proportions. Thus, an individual with four friends, two of whom are outgroupers will get the same score as an individual with 10 friends, five of whom are outgroupers, although the latter does have in fact more outgrouper friends (Laurence, 2011). On the other hand, even if we know the exact number of friends whom the respondent has or the exact number of outgroupers with whom he or she interacts, it is unclear whether an increase by one more will be substantial and meaningful. One approach is to consider the contact variable in a zero-sum relationship – a dichotomy between those who do have some form of contact and those who have very little (Laurence, 2011).

Another important form of contact is through membership in associations. Not all individuals in our surveys participate in organizations. When they do it tends to be in organizations with predominantly ingrouper membership. Therefore, 50 percent penetration of outgrouper members in clubs and organizations can be viewed as quite high. We consider contact in organizations in which more than half of the members are ingroupers as a cut-off point for bonding contact.

We adopt a similar approach to bridging contact. With the MCDS data, we have a range of items that refer to actual mixing, socializing and brief encounters with outgroupers (see Table 1). These variables correlate very highly and we build a bridging contact latent variable with eigenvalue of 1.25. The median of this latent contact variable is −0.181 for white British, and for minorities it is 0.167. The higher median among minorities is due to the greater number of outgroup contacts reported by minorities than vice versa (in EMBES, the median is three outgrouper friends). We do not want to be influenced by outliers – those individuals with very little or very high levels of contact; hence the use of median. We focus on people with low bridging contacts – the theory predicts that diversity will pose a threat particularly for this subset of the general population (Allport, 1954) and will lead to a rise in prejudice and negative sentiments. However, there is the problem of endogeneity – people with high levels of bridging contact may have chosen to live in very diverse areas and vice versa. This issue can be addressed properly only through panel data, which unfortunately we do not have. A consolation is that we also capture less endogenous forms of contact such as brief encounters; and friendship networks might be formed at a wide variety of places other than neighbourhoods, for example at work. To achieve comparability with EMBES 2010 in which we only capture friendship networks with outgroupers (an ordinal variable), we use a dichotomous measure as with the MCDS based on median number of outgrouper friends.

Models

Our response variables are dichotomous (we have tried different recodes without substantive difference to the results); therefore we use the logit link. In addition, we take into account that several individuals belong to the same neighbourhood unit by correcting our standard errors for the clustering of the residuals. Observations for which the response ‘No answer’ has been given for any of our dependent or independent variables have been deleted from the sample.

With both datasets,³ three sets of models for each outcome have been developed. The first is a baseline model which tests Hypothesis 1 with controls for individual socio-economic predictors; the second and third models are the Conflict and Contact models, which test Hypothesis 2 with the introduction of the mediating conflict and contact measures. In addition to these three sets of models we explore various interaction effects in line with Hypothesis 3.

Interpretation of Effects

We use average marginal effects, computed at different values of the X variables with the average of all these values then taken – an improvement on the mean-based approach, which in the case of dummy variables uses intermediary non-existent values. For continuous variables, the approach we adopt is to measure the effect of the change from the 10th percentile to the 90th percentile of our continuous explanatory variables (that is, diversity, IMD and age) averaged across different values of the other explanatory variables to avoid sensitivity to the extreme values at the tails of the distributions (Cameron and Trivedi, 2009).

Results

White British

In line with previous research, in our base model (Table 2 and Figure 1) we observe that, with an increase in ethnic fractionalization, both generalized and neighbourhood trust decrease. The relationship with willingness to help neighbours, although negative, is not significant. However, compared to diversity, the respondent’s socio-economic characteristics are a stronger predictor of hunkering down. Holding a degree in comparison with having no education increases the probability to be trusting in general by 27 percentage points and increases neighbour trust by 12 percentage points. In contrast, moving from a homogeneous white British neighbourhood with few minority residents (10th percentile) to a diverse one with many minority residents (90th percentile) decreases this likelihood by five percentage points for both generalized and neighbourhood trust. This inclines us to think that, considering solely the baseline models, there is some negative but modest impact of increasing diversity among white British. Is co-ethnic identification similarly affected by diversity? Remember, we hypothesized that diversity will increase identification with ingroupers in response to the pressures of growing heterogeneity. In contrast to what we predicted, in our baseline and conflict models, an increase in the percentage of non-white British in the area reduces identification with co-ethnics and quite significantly so. Furthermore, for behavioural outcomes such as civic action and willingness to help neighbours, no significant association can be noted! Non-electoral forms of participation indeed do not seem to be significantly associated with changes in neighbourhood ethnic composition, and this is a pattern that holds for both white British and for minorities as we will see in the next section.

Table 2.

Models of social cohesion outcomes – white British.

	Generalized trust			Trust in neighbours			Willing to help neighbours			Weak co-ethnic identity			Civic action
	Base	Conflict	Contact	Base	Conflict	Contact	Base	Conflict	Contact	Base	Conflict	Contact	Base	Conflict	Contact
	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)
lnage	0.052	0.081	0.053	0.084**	0.097**	0.084**	−0.051	−0.046	−0.051	−0.151**	−0.137**	−0.147**	0.135**	0.140**	0.129**
	(0.056)	(0.059)	(0.058)	(0.037)	(0.037)	(0.036)	(0.036)	(0.035)	(0.034)	(0.050)	(0.050)	(0.051)	(0.042)	(0.043)	(0.041)
Rent from Council (ref: Own)	−0.142**	−0.137**	−0.140**	−0.161***	−0.159***	−0.169***	−0.149**	−0.134**	−0.160**	0.028	0.036	0.014	−0.025	−0.027	−0.013
	(0.054)	(0.052)	(0.054)	(0.048)	(0.047)	(0.050)	(0.055)	(0.050)	(0.056)	(0.051)	(0.053)	(0.050)	(0.034)	(0.034)	(0.034)
Rent privately	−0.031	−0.029	−0.020	−0.033	−0.032	−0.030	−0.127**	−0.116**	−0.119**	−0.058	−0.055	−0.054	−0.069*	−0.071**	−0.059
	(0.057)	(0.058)	(0.058)	(0.045)	(0.046)	(0.045)	(0.054)	(0.052)	(0.054)	(0.048)	(0.049)	(0.049)	(0.037)	(0.036)	(0.038)
Other	0.066	0.009	0.045	0.079	0.060	0.067	0.027	0.002	0.008	0.097	0.068	0.066	0.101	0.091	0.089
	(0.146)	(0.139)	(0.148)	(0.080)	(0.094)	(0.087)	(0.116)	(0.130)	(0.128)	(0.154)	(0.154)	(0.146)	(0.142)	(0.136)	(0.137)
High school (ref: Basic)	0.103*	0.099*	0.081	0.020	0.020	0.009	0.013	0.008	−0.002	0.047	0.045	0.024	0.097**	0.099**	0.079*
	(0.053)	(0.055)	(0.054)	(0.036)	(0.036)	(0.036)	(0.031)	(0.030)	(0.030)	(0.051)	(0.053)	(0.051)	(0.046)	(0.046)	(0.044)
Degree	0.274***	0.213***	0.242***	0.120***	0.096**	0.106**	0.061**	0.033	0.041	0.147**	0.111*	0.121**	0.219***	0.213***	0.181***
	(0.056)	(0.060)	(0.057)	(0.036)	(0.038)	(0.037)	(0.030)	(0.033)	(0.030)	(0.057)	(0.058)	(0.057)	(0.053)	(0.055)	(0.051)
Ethnic fractionalization	−0.048*	−0.032	−0.077	−0.050*	−0.039	−0.063*	−0.022	−0.006	−0.046	0.155***	0.166***	0.119**	0.050	0.048	0.076
LnImd		−0.012			−0.002			−0.052			0.000			0.028
Fear of crime (MCDS)		−0.261***			−0.103***			−0.078**			−0.140***			−0.039
		(0.036)			(0.031)			(0.029)			(0.031)			(0.030)
Bonding contact			0.107**			0.030			0.040			0.030			0.115***
			(0.037)			(0.027)			(0.027)			(0.034)			(0.028)
Bridging contact			0.079*			0.061**			0.092***			0.127***			0.039
			(0.043)			(0.027)			(0.025)			(0.038)			(0.030)
Pseudo R-Squared	0.057	0.108	0.069	0.099	0.115	0.106	0.044	0.061	0.066	0.046	0.063	0.058	0.073	0.076	0.097
chi2	49.605	89.887	61.492	72.423	73.326	70.293	31.750	39.849	56.489	42.401	67.176	53.556	50.852	57.102	71.156
N	809	809	809	809	809	809	809	809	809	809	809	809	809	809	809

Note: * p<0.10 ** p<0.05 *** p<0.01. Average Marginal Effects (ame) and se are reported. For p90–p10, the difference is reported and the significance corresponds to the coefficient for that predictor. All models control for lnage, gender, marital status (and length of residence in the MCDS data). Source: MCDS.

Figure 1.

White British: Summary of Associations between Model Outcomes and Predictors of interest.

We should mention that we observe some interesting relationships among the socio-economic predictors. Social renting versus owning a home brings a dip in generalized trust, neighbourhood trust and willingness to help neighbours.⁴ If white British respondents rent from their local council or housing association, they appear to be less trusting (private renters are also less willing to help neighbours) – suggestive perhaps of marginalization and alienation from the larger community.

Still, one contribution of this article lies in going a step further from these frequently observed baseline relationships and in trying to distinguish between alternative ‘conflict’ or ‘contact’ explanations. When our conflict measures are introduced the observed association between growing diversity and declining social trust loses its 10 percent significance. In support of our research hypotheses, fear of crime has a very significant and distinctive negative association with four of the outcome measures – if the white British respondents in our sample are worried about crime, this reduces their probability of trusting in general by 26 percentage points, their trust in neighbours by 10 percentage points, willingness to help neighbours by eight points, and increases identification with white co-ethnics by 14 points. The effect of area deprivation, the other neighbourhood ‘conflict’ suspect is in fact insignificant.

Unfortunately, we do not know if the outlook on crime has a certain racial component to it as Quillian (1995) claims – this demands different research design and would be an interesting future line of research. So far, we can say however that white British respondents hunker down most in contexts in which they feel threatened and less safe.

What about contact? Model 3 shows that contact mitigates the negative effect of diversity in terms of generalized trust; it also decreases substantially the probability of strong co-ethnic identification. Nevertheless, the positive influence of contacts with outgroupers does not outweigh the negative association between growing diversity and neighbourhood trust – a result replicating the findings of Stolle et al. (2008) and Laurence (2011) and in line with our Hypothesis 2. Otherwise, bridging contact has a strong positive relationship with our outcomes. The marginal effect of bonding contact is also positive but significant only in the generalized trust and civic action models. This is not surprising as numerous studies summarized in Putnam (2000) observe a strong relationship between willingness to act and associational involvement with ingroupers.

Is co-ethnic identification similarly affected by diversity? Remember, we hypothesized that diversity would increase identification with ingroupers in response to the pressures of growing heterogeneity. In contrast to what we predicted, in our baseline and conflict models, an increase in the percentage of non-white British reduces identification with co-ethnics and quite significantly so. In our contact model, a four point decrease in the marginal effect of diversity is witnessed (from 16 to 12 points) but the coefficient remains significant. In other words contact is associated with a (partial) reduction in the likelihood of reactive ethnicity developing. Having bridging contacts leads to a reduction in the strength of identification by about 13 percentage points, similar to the marginal effect of degree-level education. In other words we find no sign of reinforcement of a symbolic white ‘stronghold’ identity with increasing diversity.

Minorities

The community spirit of minority respondents in our sample depends on the interplay of four characteristics: ethnicity, generation, conflict and bridging contacts.

We will start with ethnicity (Table 3 and Figure 2). Going through our range of different outcomes, black Caribbeans trust less in general than Indians but appear more trustful of neighbours than other ethnic groups, particularly more trustful than black Africans and Indians. Black Caribbeans are however less willing to help neighbours compared to Bangladeshis and more likely to profess identification with their own ethnic group. A black Caribbean is 18 points more likely than a Pakistani to identify strongly with own ethnic group and 21 points more likely than a Bangladeshi respondent.

Table 3.

Models of social cohesion outcomes – Minorities.

	Generalized trust			Trust in neighbours			Willing to help neighbours			Weak co-ethnic identity			Civic action
	Base	Conflict	Contact	Base	Conflict	Contact	Base	Conflict	Contact	Base	Conflict	Contact	Base	Conflict	Contact
	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)	me (se)
lnage	−0.060*	−0.063*	−0.058	0.125**	0.131**	0.126**	0.052	0.058	0.050	−0.155***	−0.149***	−0.161***	0.196***	0.195***	0.194***
	(0.036)	(0.036)	(0.036)	(0.063)	(0.063)	(0.062)	(0.051)	(0.053)	(0.050)	(0.039)	(0.040)	(0.040)	(0.042)	(0.041)	(0.042)
Rent from Council (ref: Own)	−0.055*	−0.041	−0.054*	−0.014	−0.001	−0.003	0.003	0.017	0.012	−0.043**	−0.039**	−0.042**	−0.045	−0.043	−0.042
	(0.029)	(0.029)	(0.029)	(0.064)	(0.063)	(0.063)	(0.044)	(0.042)	(0.041)	(0.019)	(0.019)	(0.018)	(0.029)	(0.028)	(0.029)
Rent privately	0.003	0.002	0.002	−0.066	−0.068	−0.057	−0.116**	−0.114**	−0.108**	−0.023	−0.022	−0.028	−0.052*	−0.049	−0.050
	(0.030)	(0.030)	(0.030)	(0.050)	(0.049)	(0.051)	(0.046)	(0.046)	(0.047)	(0.021)	(0.021)	(0.020)	(0.031)	(0.030)	(0.031)
Other	0.049	0.043	0.056	−0.096	−0.094	−0.082	0.013	0.011	0.028	0.094	0.094	0.105	0.227*	0.221*	0.231*
	(0.123)	(0.121)	(0.125)	(0.130)	(0.127)	(0.130)	(0.102)	(0.098)	(0.096)	(0.101)	(0.100)	(0.102)	(0.130)	(0.128)	(0.132)
Generation 2nd (ref:1st)	−0.068***	−0.067***	−0.073***	0.037	0.041	0.019	−0.004	0.004	−0.030	0.065***	0.066***	0.060***	0.112**	0.122***	0.106**
	(0.013)	(0.014)	(0.014)	(0.042)	(0.042)	(0.045)	(0.032)	(0.032)	(0.033)	(0.010)	(0.010)	(0.010)	(0.036)	(0.037)	(0.037)
Black African (ref: Black Caribbean)	0.023	0.023	0.024	−0.192**	−0.187**	−0.189**	0.024	0.029	0.028	0.017	0.015	0.018	−0.066**	−0.061*	−0.064**
	(0.038)	(0.038)	(0.038)	(0.091)	(0.091)	(0.091)	(0.050)	(0.048)	(0.049)	(0.033)	(0.032)	(0.032)	(0.030)	(0.031)	(0.031)
Indian	0.070**	0.052	0.069**	−0.156**	−0.134**	−0.148**	0.026	0.042	0.034	0.134***	0.126***	0.135***	−0.022	−0.014	−0.020
	(0.034)	(0.035)	(0.034)	(0.065)	(0.065)	(0.066)	(0.045)	(0.042)	(0.042)	(0.034)	(0.034)	(0.034)	(0.039)	(0.040)	(0.041)
Pakistani	0.010	0.013	0.017	−0.030	−0.028	−0.021	0.039	0.039	0.048	0.182***	0.178***	0.187***	−0.077***	−0.079***	−0.075***
	(0.037)	(0.037)	(0.037)	(0.051)	(0.051)	(0.052)	(0.038)	(0.038)	(0.037)	(0.036)	(0.036)	(0.036)	(0.023)	(0.022)	(0.023)
Bangladeshi	0.130**	0.138**	0.140**	−0.060	−0.024	−0.046	0.124***	0.142***	0.130***	0.207***	0.200***	0.217***	0.012	0.018	0.015
	(0.047)	(0.048)	(0.048)	(0.079)	(0.075)	(0.078)	(0.035)	(0.031)	(0.033)	(0.051)	(0.051)	(0.051)	(0.037)	(0.037)	(0.038)
Other	0.063	0.062	0.053	−0.207**	−0.176*	−0.205**	0.170***	0.172***	0.168***	0.356**	0.351**	0.335**	−0.030	−0.030	−0.030
	(0.052)	(0.052)	(0.052)	(0.094)	(0.099)	(0.094)	(0.031)	(0.027)	(0.028)	(0.111)	(0.111)	(0.113)	(0.053)	(0.053)	(0.053)
High school (ref: Basic)	−0.028	−0.028	−0.034	0.014	0.010	0.011	−0.051	−0.061	−0.057	0.014	0.013	0.009	0.124*	0.112*	0.123*
	(0.043)	(0.043)	(0.043)	(0.062)	(0.061)	(0.061)	(0.055)	(0.055)	(0.055)	(0.028)	(0.028)	(0.028)	(0.067)	(0.066)	(0.067)
Degree	0.072	0.068	0.064	0.061	0.050	0.062	−0.007	−0.021	−0.008	0.031	0.030	0.023	0.159**	0.143**	0.159**
	(0.044)	(0.045)	(0.045)	(0.062)	(0.061)	(0.062)	(0.053)	(0.052)	(0.052)	(0.031)	(0.032)	(0.031)	(0.056)	(0.056)	(0.057)
Ethnic fractionalization LnImd	−0.005	−0.0380.078**	−0.015	−0.090	−0.052	−0.078	−0.046	−0.006	−0.033	0.062**	0.056**	0.048*	0.000	0.016	0.004
					−0.105*			−0.118**			0.007			−0.053*
Fear of crime (MCDS) or Racial discrimination (EMBES)		−0.010 (0.023)			−0.033 (0.039)			−0.009 (0.028)			−0.023 (0.016)			0.046** (0.023)
Economic precariousness		−0.023**									0.004			0.032
		(0.011)									(0.008)			(0.068)
Bonding contact			0.006			0.104**			0.080**			−0.026			0.025
			(0.022)			(0.052)			(0.037)			(0.016)			(0.040)
Bridging contact			0.063**			0.068*			0.079**			0.066***			0.014
			(0.022)			(0.036)			(0.033)			(0.016)			(0.025)
Pseudo R-Squared chi2	0.029	0.032	0.031	0.038	0.044	0.047	0.062	0.073	0.077	0.077	0.078	0.088	0.12	0.129	0.121
	88.685	97.970	98.189	36.487	43.861	41.126	35.042	37.721	48.255	130.779	131.688	161.365	77.394	80.346	85.265
N	2450	2450	2450	694	694	694	694	694	694	2263	2263	2263	2450	2450	2450

Note: * p<0.10 ** p<0.05 *** p<0.01. Average Marginal Effects (ame) and se are reported. For p90–p10, the difference is reported and the significance corresponds to the coefficient for that predictor. All models control for lnage, gender, marital status (and length of residence in the MCDS data). Source: MCDS.

Figure 2.

Minorities: Summary of Associations between Model Outcomes and Predictors of interest.

The effect of ethnicity is thus much more pronounced than the effect of education – our most stable and positive predictor for white British – pales by comparison. Having a degree does not have any impact on generalized trust, willingness to help neighbours or co-ethnic identification among minorities. Other predictors such as renting from the local council, however, act in a similar pattern to the one observed among the white British – social renters are less trustful and seemingly more exposed to marginalization. Thus, a minority social renter is four percentage points less likely to identify strongly with Britain than a minority homeowner.

Second-generation minority members, as a whole, exemplify lower levels of generalized trust. Being born in Britain worryingly lowers the probability of trusting people in general by seven points. For other outcomes, however, we observe the positive relationship which might be expected in the second generation – for example, co-ethnic identification also declines by seven points.

Most importantly, for our study, diversity does not appear to have a significant negative effective on civicness among minorities. Only one of the relationships between diversity and our outcome measures is significant and in the one significant case diversity has a positive effect (increasing identification with Britain).

As in the previous section, conflict measures, particularly area deprivation, have a disengaging effect and – apart from the puzzling positive effect of growing deprivation on generalized trust and the more understandable increase in civicness for minority members who have experienced discrimination (perhaps trying to counteract discrimination through non-electoral forms of participation) – it is uniformly negative and significant. This goes against our expectation from realistic group threat theory that conflict would have a stronger adverse effect on majority members than on minorities.

In terms of contact, both bridging and associational bonding register a positive relationship with cohesion outcomes. Minority members who have median or high levels of bridging contact are, for example, six percentage points more likely to be trustful than those who have low contact levels; they are seven percentage points more likely to trust their neighbours and eight percentage points more willing to help them. Through bridging comes a greater sense of belonging to Britain rather than to one’s own ethnic group – there is a decline in co-ethnic identification of seven percentage points.

Interaction Effects

In order to test Hypothesis 3, we have run six interaction models for each of our five outcomes of interest. With so many models, one can expect that some interaction effects will appear significant without any substantive backing (often termed ‘false positives’). Indeed, we notice one significant result: there is an interaction effect significant at the 10 percent level between bonding, diversity and generalized trust among minority members. Thus, with an increase in bonding contacts and diversity, generalized trust decreases. As the effect is weak, we are cautious to over-interpret this result and to suggest it supports policy concerns that associational bonding among minority members brings deterioration of the civic mindedness. The rest of the interactions do not reach statistical levels of significance.

Discussion

If we were to build an imaginary profile of the most trusting and helpful white British citizens in our dataset, these would be individuals who are not social renters, are highly educated, mix with outgroupers but, importantly, live in diverse areas in which they are not worried about crime.

The profile of the minority members who resist ‘hunkering down’ tendencies is somewhat similar – they are home owners with outgrouper contacts; they are first generation especially in terms of generalized trust but second generation when it comes to their identity with Britain.

Importantly, there is no sign of reinforcement of either a symbolic white or minority ‘stronghold’ identity, or a decline in civic engagement with an increase of diversity. Conflict measures at the individual level have by far the most pervasive and consistent negative association across all our models which outweighs even the negative influence recorded on the part of neighbourhood deprivation for white British.

This article has addressed several popular conjectures in the literature on diversity and civic engagement. First of all, it has consistently shown that, even if a negative association between diversity and civic engagement is observed, this relationship is much weaker in terms of size compared to the profound effect of some individual predictors. Moreover, by including outcome measures such as co-ethnic identification and civic participation, we can safely say, first, that diversity plays a positive role in the formation of feelings that transcend the respondent’s own ethnicity, a result that is observed both among white British and minorities. Second, growing diversity does not have a negative impact on civic action – a crucial aspect of hunkering down – at all. Notably, since our measure of bridging contact in the MCDS data includes brief daily encounters with outgroupers not specific to the neighbourhood, we suggest (notwithstanding taking endogeneity problems very seriously) that bridging contacts are indeed important; minority respondents who had little contact with outgroupers in our sample were at greater risk of becoming less trusting and identifying less with Britain. However, we do not find strong evidence through interaction effects that deprivation or lack of outgrouper contacts moderates the way in which majority group and minority respondents are affected by growing diversity in their neighbourhoods. Perhaps this would change if direct measures of racial prejudice and stereotyping were included in the models but unfortunately they are not available in the present datasets.⁵

More generally, diversity seems to have broadly symmetrical effects on minorities and on the white British majority group, as do inter-group contacts. There is little evidence that diversity is ‘good’ for minorities but ‘bad’ for the white British. While the negative effects are sometimes significant for the white British but non-significant for minorities, they are substantively small, except when it comes to the positive effects on identity. The largest differences between majority and minorities are in the roles of education and fear of crime – both major drivers for the majority group, albeit in opposite directions. Diversity is clearly here to stay and our findings on the positive effects of inter-group contact suggest that we should not be pessimistic about the implications of diversity. It should certainly not be allowed to distract us from tackling more pressing issues of deprivation and fear of crime.

Biography

Neli Demireva is a Lecturer in Sociology at the University of Essex. Her research interests include social cohesion, diversity, multiculturalism and civic integration, job search and labour market inclusion.

Anthony Heath has directed numerous funded research projects, usually based on the collection and analysis of survey data. At present, Professor Heath works on ethnic inequalities in education and the labour market, and on the wider implications of ethnic diversity. He is also the principal investigator of EMBES, a study of ethnic minority political attitudes and behaviour, funded by the ESRC.